3D Reconstruction - Single Viewpoint

Aperçu

Explore the innovative world of 3D Reconstruction from a Single Viewpoint in this comprehensive course offered by Columbia University on Coursera. Delve into the fascinating process of transforming 2D images into a 3D structure, focusing on rigid scenes captured from a stationary camera.

Uncover the secrets of capturing images in such a way that each one reveals more about the scene, despite the fixed perspective.

Begin with understanding crucial radiometric concepts including light source intensity and surface reflectance, paving the way to mastering the estimation of scene properties like depth and material attributes. Engage with the intriguing challenge of shape from shading—deciphering a surface's shape based on its shading in a single image—and learn how changing illumination directions can compute surface normals through photometric stereo.

This technique yields a detailed surface normal map, crucial for reconstructing the shape of the surface.

Dive deeper with depth from defocus, leveraging the camera's limited depth of field to deduce scene structures from a handful of focused images. Moreover, discover the precision of active illumination techniques, employing projected light patterns for meticulous 3D scene reconstructions.

These methods are not just theoretical but have practical applications in factory automation for product assembly and quality inspection, expanding further into driverless cars, robotics, surveillance, medical imaging, and even movie special effects.

This course belongs to multiple categories including Robotics, Computer Vision, and Manufacturing, making it a perfect match for professionals and enthusiasts eager to explore the boundaries of 3D reconstruction and its real-world applications.

---

Enseigné par

Shree Nayar

---

Matières